← Back to Home

World Coordinate Alignment Guide

This guide explains how to define a meaningful world coordinate system after camera calibration in OpenLPT. By default, the optimizer produces an arbitrary coordinate frame. If your experiment requires a specific orientation (e.g., +Z up, +X streamwise), you can align the world frame using four reference landmarks on a known calibration object.

How It Works:

You identify four landmarks visible in at least two cameras: Center (origin), +X, +Y, and +Z. The software triangulates each landmark into 3D, computes direction vectors from the center, and fits the closest right-handed orthonormal basis via SVD. A rigid transform (rotation + translation) is then applied to all cameras, 3D points, and refraction plates — without altering intrinsic parameters or distortion coefficients.

            Best Practices:
            Use a rigid target — a physical frame, L-bracket, or
                    calibration object with clearly marked axis directions. Freehand points are
                    prone to noise.
Separate the points — axis markers should be well-spaced
                    (at least 5–10× the expected triangulation error) to produce a
                    stable basis.
Multi-camera visibility — each landmark must be visible
                    in at least 2 cameras. More cameras improve accuracy.
Keep the target stationary — do not move the axis object
                    between the frames used for detection.
State your convention — decide before alignment whether
                    your experiment uses +Z up, +Y up, or another convention, and label your
                    target accordingly.

        

Option A: Detect Axis Points in the GUI

Use the built-in detection and selection tools on the Point Detection sub-tab of wand calibration.

Step 1: Switch to Custom Axis Mode

In the Detection Settings panel, set the Axis direction radio button to Custom. This reveals the axis camera table and detection controls.

Step 2: Load Axis Images

In the axis camera table, click the first column button for each camera to load one or more images of your axis target. These images should show all four reference markers (center, +X, +Y, +Z) clearly visible.

Step 3: Detect Circles

Adjust the Radius Range and Sensitivity sliders so that the detector can find all marker circles, then click the Detect button. The system runs circle detection across all loaded axis images.

If fewer points are found than expected, widen the radius range or increase sensitivity.
Detected circles are overlaid on the image view for visual confirmation.

Step 4: Select Axis Directions

For each camera, click the Select button in the axis camera table. The image view enters selection mode and prompts you to click three points in order:

Click the detected circle corresponding to +X.
Click the detected circle corresponding to +Y.
Click the detected circle corresponding to +Z.

The Center point is inferred automatically from the remaining detected circle(s). Repeat for all cameras.

Step 5: Save Axis Data

Click the Save button to store the axis direction data. You will be prompted to save a CSV file. The data is also kept in memory for the current session.

Option B: Load Axis Points from CSV

If you already have axis-point coordinates (e.g., from a previous session or manual measurement), you can load them directly on the Calibration sub-tab.